Multiplexed Protein Quantification with BarcodedHydrogel Microparticles

David C. Appleyard, Stephen C. Chapin, and Patrick S. Doyle*

Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge,Massachusetts 02139, United States

We demonstrate the use of graphically encoded hydrogelmicroparticles for the sensitive and high-throughputmultiplexed detection of clinically relevant protein panelsin complex media. Combining established antibody cap-ture techniques with advances in both microfluidic syn-thesis and analysis, we detected 1-8 pg/mL amounts ofthree cytokines (interleuken-2, interleuken-4, and tumornecrosis factor alpha) in single and multiplexed assayswithout the need for filtration or blocking agents. A rangeof hydrogel porosities was investigated to ensure rapiddiffusion of targets and reagents into the particle as wellas to maintain the structural integrity of particles duringrinsing procedures and high-velocity microfluidic scan-ning. Covalent incorporation of capture antibodies usinga heterobifunctional poly(ethylene glycol) linker enabledone-step synthesis and functionalization of particles usingonly small amounts of valuable reagents. In addition tothe use of three separate types of single-probe particles,the flexibility of the stop-flow lithography (SFL) methodwas leveraged to spatially segregate the three probes forthe aforementioned target set on an individual encodedparticle, thereby demonstrating the feasibility of single-particle diagnostic panels. This study establishes thegel-particle platform as a versatile tool for the efficientquantification of protein targets and significantly advancesefforts to extend the advantages of both hydrogel sub-strates and particle-based arrays to the field of clinicalproteomics.

Advances in medical diagnostics and patient-tailored therapyrequire robust methods for the sensitive and rapid measurementof proteins. Such techniques have the potential to elucidate theactive processes that determine disease state, as well as themechanisms by which drug treatments achieve success. A high-throughput platform for the multiplexed quantification of medicallyrelevant proteins in complex biological samples will narrow theconsiderable gap that currently separates academic discussionsof proteomic analysis from the realities of the clinical setting.1-3

Although there has been moderate success in monitoring diseasestate by tracking the expression of a single protein, it is highlylikely that focused panels of protein biomarkers will provide the

most reliable predictions of therapeutic efficacy and the earliestwarnings of disease, even enabling a diagnosis before a patientdevelops symptoms.4-6

Traditionally, protein detection has been carried out with thetime- and labor-intensive enzyme linked immunosorbent assay(ELISA), which leverages the variety and specificity of antibodies.A common implementation is the sandwich ELISA, in which acapture antibody raised against the target protein is attached tothe surface of a microplate well. A clinical sample is thenintroduced; the target protein is allowed to bind to the captureantibody, and a second reporter antibody raised against a non-competing epitope of the target protein is added, thus forming asandwich. The reporter antibody is functionalized for fluorescentor colorimetric signaling. Because of the abundance of validatedantibody pairs available for sandwich ELISAs, this detectionscheme has been adapted for a number of platforms, includingplanar and particle arrays.7,8

The ability to correlate a reporting event to a specific targetspecies is crucial in developing a multiplexed sandwich assay. Withthe positional encoding scheme used in planar arrays, captureantibodies are spotted at specific two-dimensional locations, thusproviding a high density method to simultaneously measurethousands of targets.8 However, the fixed design, long incubationtimes, and low throughput of the format make planar arrays ill-suited for the rapid sample processing and frequent probe-setmodifications that are required for diagnostic applications. Aparticle-based multiplexing array that dopes polystyrene micro-spheres with combinations of dyes for optical encoding has beendeveloped by Luminex as an alternative format that can providehigh-throughput analysis of samples and faster target-bindingkinetics. Though adaptable, this system suffers from spectraloverlap between encoding and reporting fluorophores, limitingcoding capacity to ∼500. Moreover, large intra- and intertrialcoefficients of variation (CVs) require large numbers of theseparticles to be processed to generate high-quality measurements.9,10

* To whom correspondence should be addressed. E-mail: pdoyle@mit.edu.(1) Giljohann, D. A.; Mirkin, C. A. Nature 2009, 462, 461.(2) Petricoin, E. F.; Liotta, L. A. J. Nutr. 2003, 133, 2476S.(3) Zichi, D.; Eaton, B.; Singer, B.; Gold, L. Curr. Opin. Chem. Biol. 2008, 12,

78.

(4) De Angelis, G.; Rittenhouse, H. G.; Mikolajczyk, S. D.; Blair Shamel, L.;Semjonow, A. Rev. Urol. 2007, 9, 113.

(5) Gorelik, E.; Landsittel, D. P.; Marrangoni, A. M.; Modugno, F.; Velikokhat-naya, L.; Winans, M. T.; Bigbee, W. L.; Herberman, R. B.; Lokshin, A. E.Cancer Epidemiol., Biomarkers Prev. 2005, 14, 981.

(6) Wulfkuhle, J. D.; Liotta, L. A.; Petricoin, E. F. Nat. Rev. Cancer 2003, 3,267.

(7) de Jager, W.; te Velthuis, H.; Prakken, B. J.; Kuis, W.; Rijkers, G. T. Clin.Diagn. Lab. Immunol. 2003, 10, 133.

(8) Schweitzer, B.; Roberts, S.; Grimwade, B.; Shao, W. P.; Wang, M. J.; Fu,Q.; Shu, Q. P.; Laroche, I.; Zhou, Z. M.; Tchernev, V. T.; Christiansen, J.;Velleca, M.; Kingsmore, S. F. Nat. Biotechnol. 2002, 20, 359.

(9) Birtwell, S.; Morgan, H. Integr. Biol. (Cambridge) 2009, 1, 345.

Anal. Chem. 2011, 83, 193–199

10.1021/ac1022343 2011 American Chemical Society 193Analytical Chemistry, Vol. 83, No. 1, January 1, 2011Published on Web 12/13/2010

Several emerging encoded particle technologies are beingdeveloped with the intent of outperforming the Luminex system.9,11

Most of the effort by other groups has been devoted to expandingthe number of available codes, but they do not offer an efficientmethod to rapidly decode and quantify target binding, a deficiencythat severely limits their systems’ utility in real-world applications.Furthermore, many of the new encoded particles are fabricatedfrom standard photoresist materials such as SU8 that foul easilyand are not well-adapted to bioassays. This leads to poorsensitivity and large amounts of variability (e.g., 5 nM or 1 µg/mL sensitivity and CV ∼50% for IgG detection).9,12 High CVsare also encountered on “barcoded chips” due to limitations inthe manufacturing process.13 Besides the Luminex platform,metallic barcoded rods are the most mature technology in thisfield, but few protein immunoassays have been developed thusfar for the system, and the demonstrated limits of detection (∼100pg/mL for cytokines)14 are at least 2 orders of magnitude higherthan ELISA. A new approach to high sensitivity quantification,the digital ELISA, can detect down to 10 fg/mL for tumor necrosisfactor alpha, TNFR, yet does not offer high-throughput scanningnor multiplexing.15 The failure of these systems to provide aversatile approach to multiplexed protein quantification has slowedthe development of clinical proteomics. The ideal particle forprotein assays would feature a nonfouling and biologically inertsubstrate, a robust and inexpensive system for rapid synthesisand analysis, a large coding library, and a sensitivity of detectioncomparable to that of the leading technologies in the field.

The bulk immobilization of capture molecules within hydratedgel matrixes is fundamentally different from the spotting of probeonto the solid substrates that are used in planar microarrays andcurrent particle-based systems. Hydrogels have previously beenused in combination with planar array formats to provide apermeable 3-D scaffold for the bulk immobilization of captureantibody, thereby augmenting the loading capacity, improving thebinding kinetics, and reducing steric issues caused by rigidattachment.16-18 Encapsulation of enzymes and measurement ofa bienzymatic reaction have been demonstrated using suspensionarrays of hydrogel particles.19 The bioinert nature of poly(ethyleneglycol) (PEG) offers the added advantage of reduced nonspecific

binding, thereby eliminating the requirement for preanalysispurification steps.20

The use of barcoded hydrogel microparticles to quantify panelsof protein targets represents a significant paradigm shift inmultiplexed immunoassays. When the detection advantages of gelscaffolds are combined with the operational efficiencies of particle-based arrays, the platform described herein represents a powerfulnew tool for rapid and sensitive protein quantification that caneasily incorporate commercially available antibody pairs. Themicrofluidic stop-flow lithography (SFL) technique is used tosimultaneously synthesize, encode, and functionalize gel particles(>104/h) with a fluorescent barcoded region consisting ofunpolymerized holes and a spatially segregated multiproberegion embedded with capture antibodies (Figure 1).21 Thisarrangement allows a single fluorophore to be used, eliminatingrestrictions from spectral overlap and allowing for the use of asimple and inexpensive detection apparatus. The code is composedof an expandable series of five bits, each bit having a value of 0,1, 2, or 3. One bit is fixed to provide particle orientationinformation, thereby providing 192 unique codes which can easilybe augmented to >105 by adding more bits. A large code libraryenables multiplexed detection of numerous targets and theability to pool samples from a large patient population into asingle analysis. Because the particles are composed of bioinert

(10) Krishhan, V. V.; Khan, I. H.; Luciw, P. A. Crit. Rev. Biotechnol. 2009, 29,29.

(11) Finkel, N. H.; Lou, X.; Wang, C.; He, L. Anal. Chem. 2004, 76, 352A.(12) Broder, G. R.; Ranasinghe, R. T.; She, J. K.; Banu, S.; Birtwell, S. W.; Cavalli,

G.; Galitonov, G. S.; Holmes, D.; Martins, H. F.; Macdonald, K. F.; Neylon,C.; Zheludev, N.; Roach, P. L.; Morgan, H. Anal. Chem. 2008, 80, 1902.

(13) Fan, R.; Vermesh, O.; Srivastava, A.; Yen, B. K.; Qin, L.; Ahmad, H.; Kwong,G. A.; Liu, C. C.; Gould, J.; Hood, L.; Heath, J. R. Nat. Biotechnol. 2008,26, 1373.

(14) Brunker, S. E.; Cederquist, K. B.; Keating, C. D. Nanomedicine (London)2007, 2, 695.

(15) Rissin, D. M.; Kan, C. W.; Campbell, T. G.; Howes, S. C.; Fournier, D. R.;Song, L.; Piech, T.; Patel, P. P.; Chang, L.; Rivnak, A. J.; Ferrell, E. P.;Randall, J. D.; Provuncher, G. K.; Walt, D. R.; Duffy, D. C. Nat. Biotechnol.2010, 28, 595.

(16) Moorthy, J.; Burgess, R.; Yethiraj, A.; Beebe, D. Anal. Chem. 2007, 79,5322.

(17) Rubina, A. Y.; Kolchinsky, A.; Makarov, A. A.; Zasedatelev, A. S. Proteomics2008, 8, 817.

(18) Zubtsov, D. A.; Savvateeva, E. N.; Rubina, A. Y.; Pan’kov, S. V.; Konovalova,E. V.; Moiseeva, O. V.; Chechetkin, V. R.; Zasedatelev, A. S. Anal. Biochem.2007, 368, 205.

(19) Lee, W.; Choi, D.; Kim, J. H.; Koh, W. G. Biomed. Microdevices 2008, 10,813.

(20) Wan, J.; Thomas, M. S.; Guthrie, S.; Vullev, V. I. Ann. Biomed. Eng. 2009,37, 1190.

(21) Chapin, S. C.; Appleyard, D. C.; Pregibon, D. C.; Doyle, P. S. Angew. Chem.2010, in press.

Figure 1. Schematic of the SFL system for synthesis of a three-probe hydrogel particle formed in a microfluidic device with sixpressure-driven inlet streams. The streams are combined into a singlechannel where the individual stream widths can be adjusted bymodulating the inlet pressure. UV exposure causes polymerizationof the streams into a particle shape defined by a transparency mask.A collection tube is attached to the end of the microfluidic channel,and particles are gathered so they can be washed before storage oruse. A variety of chemistries can be loaded into each inlet stream.For the three probe particle, a fluorescent monomer (yellow) formsthe code and a blank monomer stream without fluorophore or probe(gray) is used to separate the code and detection regions. Threeprobe monomer streams with unique capture antibodes are used toallow multiplexed detection (orange, blue, and pink). Finally, anadditional blank stream (gray) is used to cap the end of the particle.The number of probe streams can be altered to make single-probeor multiple-probe particles, depending on assay requirements.

194 Analytical Chemistry, Vol. 83, No. 1, January 1, 2011

PEG, they are nonfouling and have minimal nonspecificbinding, making them favorable for assays of targets in complexbiological samples that may contain a thousand-fold to million-fold excess of background protein species.

Adjusting the ratio of the cross-linking monomer to an inertporogen provides the ability to tune hydrogel porosity andpermeability. Since the number of probe regions on the particlecan easily be modified by adding or removing stream inlets onthe microfluidic device, multiplexed detection can be establishedeither on a single particle or through the combination of multipleparticles.

In this paper, we demonstrate the ability to adapt traditionalsandwich antibody detection to a multiplexed barcoded hydrogelassay for three cytokines, two from the CD132 family, interleukin-2(IL-2) and interleukin-4 (IL-4), as well as tumor necrosis factoralpha (TNFR).4 This target panel is of high clinical value becausecytokine signaling initiates the immune response and is respon-sible for maintenance of the immune system memory. The IL-2family is associated with the development of adaptive immunity.IL-4 is directly involved in T-helper type 2 cell differentiation, andin combination with IL-2, it has been shown to regulate T-cellfate.22,23 TNFR is linked to the inflammatory response, is presentin septic shock, and can even act as a cancer therapy.24

EXPERIMENTAL SECTIONAssay Overview. The protein detection assay occurs in three

primary steps, which are detailed explicitly in the followingsections: Particle Synthesis, Protein Detection, and ParticleScanning. Briefly, particles with capture antibody probes weresynthesized in bulk and were stored, up to three months, beforeuse. During protein detection, approximately 50 particles werecombined with a sample containing the target protein andincubated for 2 h. The particles were subsequently washed andmixed with biotinylated reporter antibodies and incubated for 1 h.Another wash step removed unbound reporter antibody prior toan addition of streptavidin-conjugated fluorophore. After a 30 minincubation, excess fluorophore was removed with a final wash.Particles were immediately resuspended in a scanning buffer andloaded onto a microfluidic device for rapid readout of the barcodeand target level.

Particle Synthesis. Capture antibodies were functionalizedby incubating 10 µL of 25 µg/µL antibody with 2.5 µL of a 60µg/µL solution of 2 kDa heterobifunctional PEG linker (ACRL-PEG-SCM-2000, Laysan Bio) for 3 h at 25 °C while agitating at 60rpm. This mixture was combined in a 1:9 ratio with a 20% (v/v)poly(ethylene glycol) diacrylate (PEG-DA) monomer mixture(20% PEG-DA (MW 700 g/mol), 40% PEG (MW 200 g/mol), 5%Darocur 1173, 35% 3× Tris-EDTA, pH 8.0 (TE)) to form the probeprepolymer. Code prepolymer was formed by combining rhodamineacrylate at 0.06 mg/mL in 1× TE at 1:9 with a 35% PEG-DAmonomer mixture (35% PEG-DA, 20% PEG, 5% Darocur 1173,40% 3× TE). A 1:5 mixture of food coloring in 1× TE was added at1:9 with the 35% PEG-DA monomer solution to form the blankprepolymer.

Particles were synthesized using methods explicitly detailedin Pregibon et al.25 In brief, the prepolymer mixtures were loadedinto separate modified pipette tips and then injected into apoly(dimethylsiloxane) (PDMS) microfluidic device. A pressuredistribution system was used to generate laminar, coflowingstreams in a rectangular region of the device of width 270 µmand height 40 µm, resulting in a polymerized particle that has aheight of 37.5 µm (Figure 1A). The width of each prepolymerstream was precisely controlled by adjusting inlet pressure viarelief valves.25 Probe regions were adjusted to be 35 µm in lengthin the single- and multi-probe particles. This length was set tobalance sensitivity with scanning accuracy. The region had to belong enough to be accurately quantified during scanning by thephotomultiplier tube (PMT), yet short enough to concentratesignal from dilute samples. Buffer streams were placed betweenthe code and probe as well as at the end of the particle to limitedge effects on target diffusion, prevent accidental rhodamineincorporation in the probe strip, and reduce the probe width forincreased sensitivity. Polymerization was initiated through selec-tive UV exposure of the monomer streams using a negative maskof the barcoded particle (Figure 1). Computer automation ofpressure valves and illumination shutter enabled production of300 particles per minute using a five-particle mask. Each synthesiscycle used a flow period of 500 ms to establish streams, a stoptime of 300 ms to halt flow, a UV exposure time of 100 ms topolymerize particles with a Lumen 200 (Prior Scientific, 75%setting) with a 0.05 neutral density filter inline, and a hold periodof 150 ms for completion of polymerization.

Antibody Reagents. IL-2: capture (R&D Systems MAB602),reporter (R&D Systems BAF202), protein (R&D Systems 202-IL).TNFR: capture (R&D Systems MAB610), reporter (R&D systemsBAF210), protein (R&D Systems 210-TA). IL-4: capture (R&DSystems MAB604), reporter (R&D Systems BAF204), protein(R&D Systems 204-IL). Before use, antibodies were resuspendedin 1× phosphate buffered saline (PBS; Cellgro) with 0.1% bovineserum albumin (BSA) to either 25 µg/µL (capture antibodies) or500 ng/µL (reporter antibodies). Target proteins were resus-pended according to manufacturer’s specifications.

Incorporation Efficiency. Polyclonal antithrombin antibodies(Haematologic Technologies PAHT-S) were functionalized byincubating 10 µL of 100 µM antibody with 2.5 µL of a 60 µg/µLsolution of a 2 kDa heterobifunctional PEG linker (ACRL-PEG-SCM-2000, Laysan Bio) for 3 h at 25 °C while agitating at 60 rpm.A 5 µL aliquot of 7.67 µM Alexa-532 labeled thrombin wascombined with the functionalized antibodies and incubated anadditional hour. The antibody/thrombin mixture was combinedat a 1:9 ratio with a 20% PEG-DA monomer mixture. Theprepolymer mixture was loaded into a microfluidic channel (500µm wide by 37.5 µm high), and a cylinder-shaped plug 150 µm indiameter was polymerized. A fluorescence image of the particlewas taken, followed by a buffer exchange (100× channel volumeover 10 min) to wash away unattached antibodies. A second imagewas taken, and the relative fluorescent intensities of the plug wereused to estimate the percentage of antibody remaining covalentlyintegrated into the network.

Protein Detection. For single-plex calibration, 50 particleswere added to each filter plate well (Millipore MBVN1210)

(22) Sokol, C. L.; Barton, G. M.; Farr, A. G.; Medzhitov, R. Nat. Immunol. 2008,9, 310.

(23) Yates, J.; Rovis, F.; Mitchell, P.; Afzali, B.; Tsang, J. Y.; Garin, M.; Lechler,R. I.; Lombardi, G.; Garden, O. A. Int. Immunol. 2007, 19, 785.

(24) Kindt, T. J.; Goldsby, R. A.; Osborne, B. A.; Kuby, J. Kuby immunology, 6thed.; W.H. Freeman: New York, 2007. (25) Pregibon, D.; Doyle, P. Anal. Chem. 2009, 81, 873.

195Analytical Chemistry, Vol. 83, No. 1, January 1, 2011

containing the target protein in FBST (1× fetal bovine serum,Invitrogen F2442, with 0.05% Tween-20) and incubated at roomtemperature for 2 h under agitation (600 rpm; Figure 2). Wellswere washed with three 200 µL volumes of PBST (1× PBS with0.05% Tween-20). Reporter antibody was added at 5 ng/µL in PBSTto the wells to give a final volume of 50 µL, and the mixture wasthen incubated for 1 h at room temperature under agitation. Thewells were washed with three 200 µL volumes of PBST. Reporterantibodies bound to protein targets were fluorescently labeled byadding 50 µL of 4 ng/µL streptavidin-phycoerythrin (SAPE) inPBST to each well and incubating for 30 min at room temperaturewhile shaking. Particles received a final wash with three 200 µLvolumes of PBST and were resuspended in 50 µL of PTET (5×TE with 25% (v/v) PEG 400 and 0.05% Tween) for scanning. Formultiplexed detection using the interplex format with three single-probe particles with three unique barcodes, the same spike-inassay was used with 50 of each particle type added to every wellalong with permutations of the target proteins. The intraplex assay,with all three probes on one single particle with only one code,

required only 50 particles for each well and reporter antibodyconcentrations at 1.25 ng/µL.

Particle Imaging. Static images were taken with a cooledinterline CCD (Andor Clara) using a 0.05 s exposure time andillumination with a metal arc light source (Lumen 200). Imageswere combined and contrast adjusted using ImageJ (NIH).

Particle Scanning. Particles were scanned using a flowfocusing microfluidic device as described in detail by Chapin et.al21(Figure 3). Forty microliters of particles suspended in PTETwere loaded onto the device, flow aligned, and scanned past athin line illumination from a 100 mW 532 nm laser (DragonLasers) integrated into an inverted microscope (Zeiss AxioObserver A1) with a 20× 0.5 NA objective. Particles were scannedwithin an hour of resuspension in PTET; there was no detectablechange in reporter fluorescence during this period. Fluorescencesignal was collected with a PMT (Hamamatsu H7422-40),amplified using in-house designed circuitry, digitized, and storedon a PC using a National Instruments USB-6251 board at 600 kHz

Figure 2. Overview of the protein detection assay, showing an expanded view of the polymer network. (A) Premade particles containing singleor multiple capture antibody probes covalently linked into the polymer network are removed from storage. (B) Particles are combined withsample containing the target protein (purple pentagon) incubated, and then washed to remove unbound target and any other contaminants fromthe sample. (C) Biotinylated reporter antibody (orange) is added to the particles and incubated to form a sandwich with the protein target. Awash step is then performed to remove unbound reporter antibody. (D) Streptavidin phycoerythrin (SAPE, yellow stars) is bound to the complexesafter incubation. A final wash step removes unbound SAPE.

Figure 3. (A) Schematic of the scanning system used on a single probe barcoded particle. Actual scan data demonstrates how a particle is“read” providing the barcode identification and target levels. (B) Actual PMT data for a single-probe particle capturing IL-2 (code “00203”). Redline is the signal intensity for a 60 pg/mL sample; black line is a blank control particle incubated with no target present.

196 Analytical Chemistry, Vol. 83, No. 1, January 1, 2011

sample rate. Scan analysis (decoding and bound-target quantifica-tion) was completed using custom MATLAB routines.

RESULTS AND DISCUSSIONParticle Design and Coupling Efficiency. Exploring a range

of porosities by varying the ratio of reactive PEG-DA to PEGporogen in the prepolymer mixture allowed determination of acomposition which would allow for quick diffusion of targetproteins, antibodies, and SAPE while also maintaining structuralintegrity for rinsing and scanning procedures. Diffusion experi-ments were undertaken to estimate the penetration time for 150kDa IgG molecules as a function of diacrylate to porogen ratio(see Supporting Information). Penetration occurred in 65 s at thelowest diacrylate concentration used, 15%; however, the reducedrigidity of the particles produced less consistent scan results, dueto particle compression in the scanner and particle breakageduring filtration. Increasing the diacrylate concentration to 20%in the probe region improved stability with fewer than 4% of theparticles breaking during filtration. The penetration time was onlyincreased by 34% to almost 100 s using the 20% composition,suggesting a polymer network which would allow target diffusion.Prior work with 20% hydrogel particles used a 30 min incubationwith the 300 kDa SAPE to label captured targets.25 At less than30 kDa, the target proteins were expected to also fully penetratethe probe region in that time, but a 2 h incubation was used toensure that equilibrium between dispersed target and immobilizedprobe was achieved in our trials.

Covalent linkage of the capture antibody to the hydrogelnetwork allowed single-step synthesis of encoded particles andenabled spatial segregation of different capture antibodies on oneparticle. First, primary amines on the antibody were PEGylatedusing the N-hydroxysuccinimide (NHS) terminator on the het-erobifunctional PEG, and then, the acrylate moiety on the otherend was incorporated into the polymer network during particlesynthesis. Conjugation into the hydrogel may also have occurredthrough direct reaction of the PEG diacrylate monomer withsulfhydryl groups on the antibody.26 The specific mechanism ofincorporation was not determined; however, synthesis withoutprior PEGylation of the antibody was hindered by solubility issues.Estimation of incorporation efficiency through the measurementof fluorescent intensity retained after polymerization suggestedthat 26 ± 3% of the antibody in the prepolymer mixture wasretained in the final particle. The improved incorporation efficiencyover that of DNA probes (10%) may be attributed to the multiplesites for PEGylation or sulfhydryl reaction increasing the prob-ability of covalent linkage.25 The improved incorporation efficiencyand the microfluidic synthesis approach provided a cost-effectivemethod for synthesis of particles using a small volume of antibody.Five hundred micrograms of capture antibody would produceapproximately 60 000 particles, enough for over 1200 assays.

Detection Capabilities. Detection performance of the sand-wich assay was examined by generating standard calibrationcurves for each of the three proteins. To approximate thebackground protein content expected in clinical samples, each ofthe three targets was separately spiked into fetal bovine serum(FBS; Figure 4). No additional blocking proteins or filtration stepswere necessary, thereby reducing assay complexity and eliminat-

ing potential sources of contamination.27 Conservative incubationtimes were used to maximize the limit of detection, LOD, byapproaching equilibrium binding conditions. Signal-to-noise ratios(SNRs) were calculated to determine the limit of detection, target-level CV, and the dynamic range for each of the targets (Table1). The limit of detection was defined as the point where a fit tothe SNR as a function of target concentration reached three. Intra-run variations in signal were used to calculate the CV. Hydrogelparticles consistently exceeded the sensitivity of planar arrays,which required additional signal amplification to reach thedetection limits stated in Table 1.8 Additional filtration wasnecessary for bead arrays to match the hydrogel detection limit.

The dissociation constant, conjugation of linker, and loadingdensity of the antibodies play a direct role in determining the limit

(26) Salinas, C. N.; Anseth, K. S. Macromolecules 2008, 41, 6019.(27) de Jager, W.; Prakken, B. J.; Bijlsma, J. W. J.; Kuis, W.; Rijkers, G. T.

J. Immunol. Methods 2005, 300, 124.

Figure 4. Standard curves for IL-2, TNF-alpha, and IL-4 in FBS.Each point represents the average of 5 particles (IL-2), 20 particles(TNF-alpha), and 13 particles (IL-4). Inset: Signal to noise (S/N) plot,limit of detection was defined as the point where the extrapolatedline intersected a S/N of 3 (orange line).

Table 1. Detection Limits, Ranges, and Coefficients ofVariation for Bead Arrays, Planar Arrays, SandwichELISA, and Hydrogel Particlesa

target assay method LOD (pg/mL) CV (%) range (log10)

IL-2 bead array7,27,30 1.8D-8 9 3.5planar array8 7 NS 3ELISAA 1.7 3.4 3hydrogel particle 1.1 8.7 3

IL-4 bead array7,27,30 1.2D-8 8.9 3.5planar array8 10 35 3ELISAB 1.6 9.4 3hydrogel particle 8.4 11.4 3

TNFR bead array7,27,30 1.2D-8 7.4 3.5planar array8 10 7 3ELISAC 2.2-4.4 NS 3hydrogel particle 2.1 8 3

a LOD: limit of detection, defined as the point at which backgroundcorrected signal divided by the standard deviation of the blank sampleequaled three; CV, intrarun coefficient of variation defined as thestandard deviation of the signal divided by the mean of the signalaveraged over all concentrations tested in the standard curve in Figure2. A: Product information R&D systems QuantiGlo IL-2 Immunoassay(Q2000B); B: Product information R&D systems QuantiGlo IL-4Immunoassay (Q4000); C: Product information R&D systems Quan-tiGlo TNFRImmunoassay (QTA00B); D: After filtration of originalsamples; NS: not specified.

197Analytical Chemistry, Vol. 83, No. 1, January 1, 2011

of detection. Higher affinity antibodies increase the amount oftarget bound at equilibrium, thus improving signal strength.Similarly, linker conjugation, which alters or limits access to thebinding pocket, can directly impact affinity. The capture antibodydensity plays a direct role in determining the limit of detection. Ahigher density within the particle will drive equilibrium away fromfree target in solution and toward captured complexes, increasingthe number of bound fluorophores in the probe region. In lowconcentration samples, this improved sequestering of targetresults in enhanced signal and thus lowered limit of detection.Other methods for incorporating the capture antibodies at higherdensity as well as alternate capture molecules, like aptamers, arepotential avenues for improved detection.28 Though the CV inhydrogel and bead arrays are similar, the Luminex systemrequires many more particles to be evaluated to find a suitablepopulation for quantification. The graphical barcode and microf-luidic scanner enables quantification using fewer than 25 hydrogelparticles whereas the spectral coding of bead arrays may requirehundreds or thousands of particles.29 The signal response per unitbound target is greater for proteins than for miRNA or DNAsystems previously evaluated using hydrogel particles. This canbe attributed to the reporter antibody having multiple binding sitesfor fluorophores, as opposed to the single site of a nucleic acid.While this improves the limit of detection, it also reduces thedynamic range as the PMT detector saturates at high targetconcentrations. It should be noted that this is a limitationassociated with the particular detector used in our setup and notwith the saturation of the immobilized probes. The selection of adifferent PMT with an expanded range would correspondinglyaugment the detection capacity of the platform.

Multiplexed Detection. Multiplexed detection is essential foreffectively monitoring complex systems in clinical diagnostics orresearch applications. The SFL method enables the spatialsegregation of multiple probes on a single particle, yieldingconvenient colocalization of common target groups. Graphicalbarcoding enables the use of multiple particle types in a singlesample without the limitations imposed by spectral overlap ofencoding fluorophores with target quantification fluorophores,thereby enabling rapid probe-set modification for the investigationof a wide variety of target panels. In complex samples, streamlinedidentification of barcode and quantification of target abundanceis obtained using the microfluidic flow-focusing device. Thecurrent design can scan the contents of a 50 µL sample well in30 s, and serial scanning of a standard curve can be completed inunder an hour even with manual sample loading of the PDMSdevice. In future applications, this throughput can be expandedby one to two orders of magnitude with the implementation of anautomated liquid handling system.

Cross-reactivity of reagents is one of the primary challengesin developing a multiplexed protein assay.3 Therefore, it is vitalthat any new protein quantification platform be capable ofintegrating reagents already shown to be compatible. Thissignificantly reduces barriers to assay development and enhancesthe utility of the emerging platform for immediate application.Multiple research groups and commercial sources have assembled

validated antibody panels which could be seamlessly paired withour hydrogel technology. We were able to demonstrate simulta-neous detection of the three cytokine targets using three uniquelybarcoded particles with a single probe strip. All possible combina-tions of the three cytokines were examined to isolate and identifypotential sources of cross-reactivity (Table 2). Microfluidic scan-ning and analysis of the particles provided quick identification oftarget type and level. Recovery of the FBS spike-ins was within20% of predicted values from the calibration curve (Table 2). Mostimportantly, no significant source of cross reactivity betweenantibody pairs was encountered.

A major advantage of our hydrogel design is the ability tospatially address probe location in combination with barcodeidentification. We created particles with all three probes proximallylocated on a single hydrogel particle. This “intraplexing” reducesthe number of particles necessary for an assay, allows the creationof a single code particle for a specific marker set, and acceleratesthe reading throughput of the flow cytometer, while also extendingthe available code base to allow for higher degrees of multiplexingand patient sample pooling. Under the same conditions as theinterplex, we were able to detect all three targets on a singlebarcoded particle (Figure 5). As in the earlier trial, the limitedcross-reactivity in this intraplex assay, even at relatively highconcentrations, indicates the compatibility of the hydrogel systemwith conventional antibody pairs and panels. Expanding thenumber of probes on the particle only requires the simple additionof extra monomer streams producing up to a maximum of eightprobe regions on a 250 µm particle with limited barcoding. Thesimple scalability of intraplex particles results in efficient use ofparticle space in assays with limited sample volumes.

Future Directions. Experimentation with additional antibodypairs used in ELISA and bead-based assays will continue to expandthe available targets for multiplexed detection. Further modelingof LOD as a function of antibody loading will allow for a moretailored detection range and potentially enhanced sensitivity. Usinga microfluidic device capable of holding particles while allowing

(28) Cho, E. J.; Collett, J. R.; Szafranska, A. E.; Ellington, A. D. Anal. Chim.Acta 2006, 564, 82.

(29) Peck, D.; Crawford, E. D.; Ross, K. N.; Stegmaier, K.; Golub, T. R.; Lamb,J. Genome Biol. 2006, 7, (7): R61.

Table 2. Multiplex Detection of Three Cytokines Usingthe Interplex, Three Particle Systema

target signal (mV)A

IL-4 IL-2 TNFR IL-4 IL-2 TNFR

+ + + 927 ± 98 1271 ± 96 1471 ± 145+ + - 750 ± 161 922 ± 143 128 ± 57+ - + 666 ± 33 76 ± 15 1055 ± 77- + + -88 ± 64 1243 ± 129 1382 ± 46- - + 35 ± 40 51 ± 42 1340 ± 122- + - -144 ± 87 1105 ± 41 45 ± 11+ - - 850 ± 128 13 ± 7 70 ± 109- - - 0 ± 92 0 ± 8 0 ± 17

recovery (%)B 101.7 ± 15.9 118.0 ± 16.5 93.1 ± 16.7

a A: Signals were background corrected by subtraction of the controlparticle (-,-,-), resulting in small negative signals for some conditionswhere targets were absent. Presence of target (+) correctly corre-sponds to detected signal. Data taken from five particles for each targetunder each incubation condition. Negative values are the result ofincubations in which the target amount was below the system limit ofdetection and reflect the noise that dominates the system in the casewhere there is no detectable signal from the target. B: Recovery forspiked samples calculated as scanner measured concentration usingstandard curves in Figure 2 divided by the actual spike in concentration.Values are mean ± SD for the four multiplex samples with targetpresent. TNFR spiked at 0.4 ng/mL, IL-2 spiked at 0.12 ng/mL, andIL-4 spiked at 0.11 ng/mL.

198 Analytical Chemistry, Vol. 83, No. 1, January 1, 2011

for buffer exchange, like that described in Zhang et al., providesthe potential for integrating protein detection and scanning stepsonto a single device for improvements in throughput and reagentconsumption.31

CONCLUSIONSBarcoded hydrogel particles combined with the microfluidic

scanning system provide a high-throughput platform for multi-plexed quantification of protein abundance and need not be limitedto DNA and RNA detection. The ability to covalently incorporate

antibodies into the polymer matrix allows for the creation of single-and multi-probe particles capable of measuring clinically relevantamounts of proteins with limits of optical detection, coefficientsof variation, and dynamic ranges comparable to leading methodscapable of being implemented in a clinical environment. On-particle multiplexing demonstrates the unique capacity to spatiallyand geometrically encode particles without concern for spectraloverlap, while allowing for high-speed reading with a microfluidicflow-through scanner. We also believe this concept will allow forthe design of new assays. Use of a bioinert PEG hydrogel reducesbackground in complex samples and eliminates the need forblocking proteins or prefiltration of samples. The successfulcombination of multiplexed protein quantification with the versatilenature of SFL provides a flexible platform for clinical diagnosticsas well as tailored research applications.

ACKNOWLEDGMENTWe acknowledge support from Grant R21EB008814 from the

National Institute of Biomedical Imaging and Bioengineering,National Institutes of Health, and partial support from the RagonInstitute of MGH, MIT, and Harvard.

SUPPORTING INFORMATION AVAILABLEAdditional information as noted in text. This material is

available free of charge via the Internet at http://pubs.acs.org.

Received for review August 25, 2010. Accepted November30, 2010.

AC1022343

(30) Lagrelius, M.; Jones, P.; Franck, K.; Gaines, H. Cytokine 2006, 33, 156.(31) Zhang, H.; DeConinck, A. J.; Slimmer, S. C.; Doyle, P. S.; Lewis, J. A.; Nuzzo,

R. G. Chem.: Eur. J. 2010, in press.

Figure 5. Multiplexed detection on multiprobe particles (“intraplex”).From the top, probe strips are TNFR, IL-2, and IL-4. TNFR spiked at0.4 ng/mL, IL-2 spiked at 0.12 ng/mL, and IL-4 spiked at 0.11 ng/mL.

199Analytical Chemistry, Vol. 83, No. 1, January 1, 2011